We study the minimum mean-squared error for 2-means clustering when
the outcomes of the vector-valued random variable to be clustered are
on two spheres, that is, the surface of two touching balls of unit radius in
-dimensional
Euclidean space, and the underlying probability distribution is the normalized surface measure.
For simplicity, we only consider the asymptotics of large sample sizes and replace empirical
samples by the probability measure. The concrete question addressed here is whether a
minimizer for the mean-squared error identifies the two individual spheres as clusters. Indeed,
in dimensions
,
the minimum of the mean-squared error is achieved by a partition obtained from
a separating hyperplane tangent to both spheres at the point where they touch. In
dimension ,
however, the minimizer fails to identify the individual spheres; an optimal partition is
associated with a hyperplane that does not contain the intersection of the two spheres.
PDF Access Denied
We have not been able to recognize your IP address
18.97.14.85
as that of a subscriber to this journal.
Online access to the content of recent issues is by
subscription, or purchase of single articles.
Please contact your institution's librarian suggesting a subscription, for example by using our
journal-recommendation form.
Or, visit our
subscription page
for instructions on purchasing a subscription.